Monday, October 17, 2011

Moore from the Land of Harsh 12v busses and Itty Bitty Batteries

In this week's underdog adventure, we retrace our steps on the Aux battery and the 12volt subsystem.

We have a number of loads that need to be on all the time - parked or running. In fact, the number of these is reaching kind of alarming proportions.

The latest that slipped under the radar was David Kerzels' SAE J1772:2010 Compatible Active Vehicle Side Control Board Module This little device is $37 on eBay and is all you need, besides the inlet itself, to respond appropriately to any commercial J1772 EVSE and charge therefrom.

I love it. Not only does it do the necessary response to the EVSE to both trigger the copilot signal to start the charge process, but it also monitors the proximity switch (the little button on the charge plug) to stop charging if it is pressed. AND it provides a little single pole double throw relay you can use to detect it's state to do such things as interlock your onboard charger, disable the controller while charging, whatever.

The unit draws 20 ma, although we use the SPDT relay to also apply 12v to a string of blue LED lights ringing the billet aluminum fuel port on the Cobra. Good visual at the car that it is charging when the blue LED is on.

But the trick is, it has to have 12v ALL THE TIME, or you can't charge. No 12v to the board, no charging.

We are powering our Xantrex AH meter also with 12vdc. We've tried a number of near disastrous ways of powering this little device and the best seems to be to use an inexpensive (usually $10-$12) DC-DC converter module to take 12vdc from the car, isolate it through the 12v/12v converter, and power the Xantrex with it. Of course, if we are to account for IN amp hours while charging, we have to have it on all the time. Since I can't read it without the backlight, that's on all the time as well. It's a bit of a 24x7 load on the 12v system.

Then there's the ZevaII Fuel Gage Driver. I spend a lot of time exasperated with this device as it is tricky to install and calibrate. But I do love it. It does two things I like. First, it runs an ordinary fuel gage by counting ampere hours. You can use this device to fairly accurately monitor your LiFePo4 pack's state of charge using a familiar and already standard face - an ordinary fuel gage. We have a set of Speedhut Cobra gages for the Cobra and it comes with a fuel gage of course. Might as well use it. Probably NOT as accurate as the Xantrex, but easy to use and everybody knows what it means.

This device also outputs a pulse string representing instantaneous current. We can display this in hundreds of amps on our tachometer from Speedhut. Actually, we installed a small selector switch in the dash that let's us connect this output or the output of a RechargeCar magnetic shaft pickup to our tachometer. In this way, we can read RPM or battery amperes on the tachometer.

This is kind of cool A dancing set of digital numbers for instantaneous current is not really VERY informative. We're changing the current faster than the digital device can sample. But if we put the same information on a large analog needle, it is all different. We get a rougher indication of value, but a kind of more integrated "trend" visual that shows us the current going up and going down in response to our throttle inputs.

Finally, the ZivaII let's us set an alarm level and provides a switched ground output when we reach some low level of charge. We can hook that up to a warning light or power a relay with it.

But to do all that, it is ON for 24x7. In fact, if you remove power from the device, it resets your AH counter and your fuel gage will show full when it's actually had a number of amp hours used This is a worse situation than having no fuel gage at all.

So we tried using an aux battery to the ignition switch, and that voltage to in turn run a contactor that turned on our main DC-DC converter. This quelled the complaints from the Netgain Controls Warp Drive Industrial, but caused endless other problems. Most notably, we kept running down the battery when we charged overnight.

We added a manual bypass switch to bring up the DC-DC converter and recharge our dead battery. But the whole thing was a nightmare and we were going to have a totally destroyed battery within days.

We have tried an aux battery in the past - always to some bad end or other. We've had numerous people point out that if we lose our DC-DC converter, our car won't operate. Duh. What we've found is if we lose our DC to DC converter, we don't know it at first because we have an aux battery. But within a few minutes, the battery runs down anyway and our car won't operate. Plenty of weight, an inherent maintenance item, and same effect.

We're also warned of dire safety issues as a sudden shut down will cause us to lose power brakes and power steering and all manner of ills suddenly while hurtling down the freeway. Same answer really. But most of our cars don't have power steering, power brakes and so forth anyway. For the ones that do, for example the 2009 Mini Cooper Clubman Electric, we have a different solution - redundant DC-DC converters. Indeed the design of the Mini's startup sequence requires 12v of course, and we use one converter to bring it up, and when up it engages a second converter as well. When we shut down, we disconnect the big one, but leave the small one running.

The original problem on the eCobra had to do with the Warp Drive Industrial controller. It threw a series of errors when we shut OFF the controller at the end of the drive. This was very peculiar. But the controller stored the errors and so wouldn't operate the next time we fired it up. We learned to clear the errors with the Interface Module, but once we have all this working, we don't see really using this Interface Module day to day. A great troubleshooting device, it can't count Amp hours and doesn't really add much operationally. Certainly having to clear these errors prior to driving wasn't happening.

Ryan Bohm insisted this was caused by noise on the 12v system and wanted us to scope it. I have a problem with all that. What noise are we looking for? There is noise. And there is noise. And in fact, we find the 12v system in a car a very noisy place, and one of the primary culprits is in rather circular fashion - the controller. So I'm vaguely disinterested in noise in general. Assume we are going to have it in an automotive environment.

In this case, I had another problem with the concept. The controller worked fine - ONCE we had cleared the errors. Indeed, the errors only occurred when we were shutting DOWN the system. If the controller has an issue with "noise' in general on the 12v system, why didn't it throw errors while I'm driving down the street?

Our 12v system does have an unusual number of inductive loads on it. We have a prius pump, for the controller cooling system, a Derale heat exchanger fan of actually some size, again for controller cooling, and an XSTurbos turbocharger we are using as a cooling blower for the 11HV motor. At 435 cubic feet per minute, this fan draws 4.6 amps at 12v.

The windings of a motor are an inductor. An inductor resists changes in current. Current through the windings causes the expansion of a magnetic field around the windings. This field stores energy. When you cut off the current, the field collapses. This collapse induces a current in the same direction. And this can cause relatively huge voltage spikes. The spikes are hard to see on an oscilloscope because while they can be surprisingly high in amplitude, they are very transient. They only appear for a handful of milliseconds. I happened to have a very cool very fast 600v 100 amp diode laying around - about 10,000 X overkill for this test. But we hooked it up between our 12v bus and ground.

It had NO effect.

Ryan Bohm was able to duplicate the problem as he had a Derale heat exchanger as well. What he found was something very different. The fans have a bit of inertia in them. When you shut them off, they keep spinning for a second or two. And when they do, they act as generators. They were maintaining 4-5 volts, quickly decaying of course, on the 12 volt line.

The problem is that the ON digital input to the controller uses this 12v to signal the controller that we are indeed on. So we have shut DOWN our 12v supply to the controller, but the fan is producing sufficient voltage that the ON input is high - for a second or two. This causes the errors to be noted and stored.

We tested this by putting a diode in series with the fan. When it is shut off, the fan does generate, but the diode blocks the feedback into the line.

That pretty much fixes the problem, but as the Derale uses six or seven amps of current, I did not want a semiconductor in that line heating up and inevitably becoming a failure item.

As a workaround, we connected the fans with a relay and used the ignition voltage to energize the relay. In this way, when we turn the ignition off, the fans are simply disconnected physically fro the system. The 3-5v cannot reach the ON input, and the errors do not occur.

This was an interesting problem. But it had us running in circles for days. And it is probably beyond our concept that ANYONE can convert a car to electric drive. in this case, anyone can't, and indeed Jack could not for some period of time. So the "fix" is for the controller to be fixed where this is not an issue.

It does bring up the concept that all such loads ought to be on separate relays with separate fuses. Not a bad practice actually, and you will see this in most modern automotive fuse blocks. Lots of fuses, and lots of relays. There's a reason. But the controller should not depend on that to operate. Put a relay inside the controller if you like.

We also devoted a bit of time to a discussion of the Tesla/Panasonic connection. Tesla makes all this needlessly secretive and confusing with their constant claims of proprietary madness. But that's mostly illusion and press puffery. Panasonic did invest $30 million in Tesla and owns 2% of the stock in the company. And they are working together. But the original concept was for Tesla to use off the shelf cells that are produced in the millions for laptops, cell phones, and flashlights. With the acquisition of Sanyo, Panasonic is the largest manufacturer of those in the world, BYD perhaps excepted. on raw numbers.

In March of 2012, Panasonic begins the production run of a brazillion of their new NNP 3.4Ah cell. This is there New Nickel Platform. It's actually a Lithium Nickel Cobalt Aluminum Oxide cathode material with a carbon anode. And the thing carries 12.24 wH of energy in a 46 gram package. This is something like 266 wH per kilogram. By contrast, the CALB or Winston type cells are more like 109 wH/kg in their optimum size.

This is very nearly two and a half times the energy by weight. And if our 444 lbs of CALB cells were these new Panasonic's instead, we'd be seeing 266 miles instead of 109 in Speedster Duh. As I've actually driven Duh pretty flat at 110 miles, I'm onboard.

Ergo the 300 mile Tesla Model S.

We also learned a bit about cycle life in these cells which is very encouraging.

Cycle life to the industry "traditional" 80% of original capacity is somewhere in the 500-800 cycle range. Not good. But if you look to 70% as the mark, it gets much better as the deterioration curve flattens out in a very unusual fashion. You're looking at well over 2000 cycles to that level. And 70% of 300 is still 210 miles.

These are not here yet. But I do not classify them as unobtanium per se. They have good prospects for being available subsequent to March 2012. Tesla has contracted for enough of these cells to do 80,000 cars over four years. That's 640 million cells. As it will also be the highest energy density of any 18650 form factor cell, it will undoubtedly be popular in a number of other applications - flashlights if no where else. And so we have a truly MASS market battery cell here. That brings in economies of scale. That's the game we need.

I don't relish making modules of all these little cells. But 2.5x energy at potentially a LOWER price within a few years from what we're now paying for CALB/Winsston cells could be a play.

In a much larger sense, there is a perception, espoused by Elon Musk himself, that the Moore's law of PC speed and bandwidth only applies to batteries in a rather sedate form of 8% per year or so. I've never completely bought into this. The creeping advance were in a product that had zero market and no cash flow. As soon as you add oxygen in the form of ducats coming in the door in substantial numbers, innovation in batteries is not that hard. There is tons of cool science laying around with the usual problems of engineering to a product level, but there simply has been no oxygen (money) to drive the productization. Panasonic has already announced plans for 2013 to bump this very 3.4Ah cell to 4.0Ah by using a silicon alloy anode in place of the carbon anode in this cell, for example. But if a fire ensued in the battery market, there are plenty of players and plenty of advances to come. I think Moore's law is alive an dwell in Batteryville. And a car with a 1000 mile range is not inherently a preposterous notion.


  1. Jack,

    How about doing a segment on the setup and calibration of the Zeva II with the Speedhut fuel gauge? I have fiddled with mine and have yet to break the code. A little wisdom from the EVTV shop would help a lot!


  2. By the way, I do have the Zeva II ammeter output reading on my tach, and I agree, that's really useful.

  3. Jack, regarding the switched Netgain controller input.
    Why didn't you just use the double pole relay with one pole connected to the controller input and the other one to the rest of your 12-volt devices?
    Why do you want to connect some of the devices to the separate relay, and leave the rest connected to the controller?
    If your controller would be alone on one of the relay outputs, nothing could go wrong.

  4. Has anybody here checked out the "Cycle Analyst" for ev use?

    Seems to be a fairly well endowed and tasty bit of kit for very fair money.

  5. @Andyj: Many of us use the CycleAnalyst. It is the use in EVs which prompted the large display version and the production of high voltage versions.

  6. Fred:

    We'll just do it. I agree, it's a mess of a procedure. But I think I have it decoded.


  7. to me always-on parasitic loads are not acceptable unless they are actually necessary. in this case I don't think any of them are.
    I would simply add a momentary button to supply the J1772 board so it can begin charging and for the duraction of the charging switch on the DCDC to supply the j1772 board and the Ah counter.
    running the DCDC 24/7 seems like a failure source to me as it ages maybe 50 times faster than otherwise.

  8. Dan,
    More often than not. Electronics tend to experience more failures when mechanically switched on and off.
    Thanks Giz,
    Just read your blog.
    Why I mentioned the Cycle analyst, is simply because it does not require a constant DC source to remember all its settings and it works off a very large voltage range of 10~150V or 20~350V.

    I was taken aback how handy and complete this device is and at such a reasonable price. A pity it doesn't drive posh analogue clocks.

  9. "Why I mentioned the Cycle analyst, is simply because it does not require a constant DC source to remember all its settings and it works off a very large voltage range of 10~150V or 20~350V."

    Which also means it doesn't put an imbalance on the pack! It has a small capacitor such that when the power is cut off to it there is just enough energy for the unit to store the current settings in nonvolatile RAM before powering down. I've wondered why the ZEVA II doesn't do something similar.

    Also, support is awesome!

  10. Hi Jack,

    Congratulations for solving the controller error problem. So, are you throwing out the aux battery?
    The panasonic battery is indeed a very exciting prospect.


  11. The aux battery is history.

    The Panasonic battery is a demonstration of what is coming in batteries. Winston Chung is now alluding to a much stronger battery as well.

    I think we're entering a good period for batteries. The cells you all have bought at very high prices has provided a bit of oxygen in the room for battery advances.

    Jack Rickard

  12. John,
    The aux batt was removed as a very bad idea. It's now hopefully fixed and you can find it here:
    4 minutes in for 20 minutes.

  13. oops! Sorry. Late out of bed....

    I'd like 1100 of those cells in 50 tubes of 1.43metres thank you. Or maybe double that. People would be asking where is the battery pack...

  14. Hi Jack,

    Do you think there is any merit in doing a roll down test?

    The following instructable and associated comments should be able to give you a breakdown of the energy required to roll and energy required to overcome drag for your cars.

    I like one of the comments suggestions in using a camera and a graphics app to determine frontal area.



  15. I am basically planning on doing a coast down test on the eCobra and probably all of them before it is over. I don't know that we'll take hit as far as computing rolling resistance vs aerodynamic drag using the classic formula. That gets to be a little more complicated than quite covered by your friend in the instruct able.

    But we should be able to produce the curve and perhaps some figure of merit representing it that would give us a way to compare one car to another and be able to detect the effect of improvements in a somewhat more facile way than the Soap Box Derby. The problem of course is finding a truly flat area of that length, with no traffic, and no wind. But I'm looking actually.

    But I agree a coast down test from above 70 down to zero would be much more exhaustive. And we're looking at it.

    Jack Rickard

  16. I bought a Cycle Analyst a year ago. I was seriously annoyed with it. I had trouble actually getting it calibrated and it would not measure at different amp levels consistently. Worse, the guy didn't deb ounce the keys and so it was very frustrating to setup.

    I guess I think someone writing software for such a thing that hasn't mastered a very basic such as button deb ounce probably isn't doing anything good for me. I laid it aside and we've never mentioned it on the show.

    On your recommendation, we'll try to acquire the latest version and try again. Things get better over time.

    Jack RIckard

  17. Giz seems to be happy with his. Maybe a simple update will clear?

    I was fired up, all set to grab a copy of flash to make a pretty set of analogue clocks but they don't do it for Linux. Grrr!

    Promise, will be all ears and eyes, meooww.

  18. Jack,

    The button which came with my CA wasn't very good and Justin, the designer builder, said it was just to get going any way and recommended I get the button style I wanted to match my install. I just picked up a couple of buttons from Radio Shack and they worked just find and I haven't had any issue with button bounce. Mine is version 2.2. IIRC, the most recent one is 2.3 and comes in its own case rather than a bare board with display. It was really easy to calibrate the shunt, voltage, wheel size, etc. It definitely is simpler if both buttons are installed as no power cycling is needed to get into the setup mode.

    David D. Nelson

  19. David:

    I wasn't complaining about the quality of his buttons. With micro controllers, you can easily wire up a button as a digital input. But when you see a change, you must make SOFTWARE provisions to blank out any further inputs from that button for some tiny period of time to keep from getting multiple inputs from a single button press. This is because my fat thumb doesn't operate well at the millisecond level.

    This programming technique is called DEBOUNCE and it is very basic to micro controller software design. If he doesn't know how to do that, how good is the rest of the software likely to be?


  20. Jack, if you are taking the plunge. I hope this will be of interest for you.

    Some Cycle Analyst clocks as an iphone/ipad app.

    His Source code:

  21. I'm aware of the debounce technique. I do not know what, if any, debounce delay is programmed in but it is possible that with the poor button that the bounce time was longer than the debounce delay. In any case, one mistake doesn't make everything else done bad or poor. Since a button change fixed the problem and I've had no other problems with the unit maybe the delay just isn't long enough.

    @Andy: thanks for the links.

  22. Do you have any information on how many watts these new Panasonic cells can sustain? Energy density is only half the story in an automotive application, we also need to know the power density.

  23. Gizmo:

    You're probably right. But usually if the hot and sour soup is no good, the szechuan chicken isn't going to be any better. I was frustrated with the device after about 45 minutes and just tossed it saide. So many toys. So little time.

    Escape Velocity. The Panasonics are high energy, but low power. They dance around it in the specs and descriptions, but it implies a 1C output. Maybe one reference to 1.5C in a chart somewhere.

    Not very encouraging.

    You can have density, power, or availability. Pick any two.

    Jack RIckard

  24. Hi Jack.

    I have been measuring self-discharge or lack of same on my TS90AHA. I could add about 1Ah to two cells I had sitting fully charged on the bench top for a year. This translates to about 0% self-discharge in my book and are confirming your numbers.
    I did a full discharge/charge cycle and could see a 5Ah permanent capacity decrease. Did you check the capacity on your old cells?
    I then took two cells from my car, which had been doing 7000 miles in all weather (-15 degr. C - 25 degr. C) on my more or less daily commute which take the battery to 60% charge twice a day.
    I could measure about the same 5Ah decrease in capacity - so from the numbers, I could not tell whether the cells had been sitting on a bench top or in the car.


  25. Interesting. These are hard measurements to get. I'll offer this. Do a full charge/discharge on the same cell that you saw the 5Ah decrease on. Actually either one.

    I have found it very difficult to get the same number twice. And it usually gets BETTER not worse.

    Use exactly the same voltage and current levels and see what happens.

    Jack Rickard

  26. "Do a full charge/discharge on the same cell that you saw the 5Ah decrease on."
    I saw the same approx. 5Ah drop on all four cells (two from the bench top and two from the car). Since I buddy paired the cells I have one from each batch in each pair. The two on the bench top was not paired, but the two from the car was tie together during use, but not during measurement.

    The charge/discharge capacity is very close on three of the four cells but the last of them has a 2Ah difference when I compare the charge/discharge capacity - I will measure that one again. This single cell also shows a larger dip in capacity. It is the high capacity one of the buddy pair from the car.

    I am using calibrated equipment, but manual set currents and voltage levels, they are however so close so that should not (tm) introduce any significant errors. But when I am done, I will put the data sets on my web site.


  27. This happens to me a lot. Using your calibrated equipment, you know that it SHOULD NOT introduce any significant errors. When you get a different result, it is interesting to try to explain why. After awhile, it will dawn on you that the expected result is not always the one you get.

    I think the heart of learning about these cells is learning to look at the meter and read what's there instead of what you expect to be there. It has got me over and over again.

    Jack Rickard

  28. I know - voltages and currents has been triple checked with other calibrated and different equipment - they agreed with in 0.5%. That should be good enough for government work ;-)

    But the second of the four cells are discharging as I write this.


  29. Discharge data for the cells from the work bench (1+ year at full charge):
    F18315 July 2010: Initial 92.4 Ah
    October 2011: 87.2 Ah
    F20478 July 2010: Initial 101.6 Ah
    October 2011: 94.5 Ah

    Discharge data for the cells from the car:
    F18333 July 2010: Initial 93.8 Ah
    October 2011 1'st run: 90.7 Ah
    2'nd run: 90.4 Ah
    F20474 July 2010: Initial 100.2 Ah
    October 2011 1'st run: 94.7 Ah
    2'nd run: 94.5 Ah

    In all cases the cells was charged with 30A until the cell voltages was 4.0V and the held at 4.0V until the current was 4.5A. So not quite 0.3C or 0.05C, but close and same conditions in all three cases. I used the same procedure in order not to change to many parameters. I am not quite sure if I should use nominal or real capacity, but I do not think it matters that much as long as I am using the same procedure each time.


  30. Forgot to mention that the cells was discharged with 30A until they reached 2.8V


  31. Hello Jack,
    One gentleman wrote a little while back about encoded remote key switching you can fit into a car. I saw the site and thought it was extremely expensive for what it was.

    Someone handed me this link:

    Two 10A relays on 12V from two buttons. At a horrendous price of $8.64 +postage to the US(?).

    Might buy their 100 off packets of 5mm, 15,000Mcd LED's to have a go making up my proposed 72V headlight circuit. Considering they drop 3.3V per LED.